The ability of a new continuous-wave Doppler esophageal probe to measure cardiac output noninvasively during surgery under general anesthesia was tested and compared with simultaneously measured thermodilution cardiac output. A Doppler computer, calibrated for the aortic diameter and the transcutaneously measured cardiac output from the suprasternal notch, computed the Doppler cardiac output from the descending aortic blood flow velocity signal. A total of 246 paired Doppler cardiac output and thermodilution cardiac output measurements were made in 14 patients during surgery. The average thermodilution cardiac output was 5.90 +/- 3.27 (standard deviation) liters/min (range 1.20 to 19.18); the average Doppler cardiac output was 6.21 +/- 4.0 liters/min (range 2.30 to 28.20). The difference between the cardiac output measured by the 2 techniques was 1.38 +/- 2.2 liters/min (range 0.04 to 16.8). Two to 5 cardiac output measurements were averaged and arranged into "time periods." The average standard deviations for thermodilution and Doppler cardiac outputs within each time period were 0.64 and 0.47 liters/min, respectively. There was a correlation between the 2 measurements over a range of cardiac output values (r = 0.76, Doppler cardiac output = 0.93 x thermodilution cardiac output +0.7, standard error of the estimate = 1.76). Reproducible measurements of Doppler cardiac output were obtained during intraobserver (mean difference 0.64 +/- 0.52 liter/min) and interobserver (mean difference 0.41 +/- 0.36 liter/min) studies (n = 8). Cardiac output measurement by the Doppler esophageal probe could be used for hemodynamic monitoring during surgery in selected patients with cardiopulmonary disease.